Just to further clarify I don't think this is an issue with the hardware offloading this is just how traffic works on layer2 domains as well as how Mikrotik handles the traffic flow.
Obviously this is directed for OP rather than MKX even though this sort of looks like a reply?
So to understand why this is not working you will need to understand how same subnet traffic works, to put it simply we dont route we arp. If you have a look at your routing table you will notice some dynamic routes with DAC (dynamic, active, connected) that get put into the device automatically when installing a IP address and if you look at the gateway you will notice that it is an interface not an ip address.
To put it simply when the router seeing this it will know that it has to use ARP to find out how owns that IP address rather than just being forwarded on to a router that has a path to get to that address.
So when you try to go to the same subnet your device will send out a broadcast and say hey who has x.x.x.x and everyone on the same layer 2 domain(do not get that confused with same broadcast domain as that is to do with IP addresses/subnets not layer 2 connectivity in general) from there is someone has the address they will reply and say yep that's mine here is my mac address if ya wanna chat then they both put each others mac address to ip translation in their arp table for a little while.
Thats as simple as I can make it, if you would like something more in depth let me know and if I have time ill write up something a bit more 'official' but the important thing here is to know that we 'dont route' to same subnet devices we use ARP
ok now that we know we are not going though any of the routing forwarding tables we can get an idea of what the traffic is doing, now we need to pair this up with the packet flow, if you look at:
https://wiki.mikrotik.com/wiki/Manual:Packet_Flow
This explains how the traffic moves around the router and what processes will be hit.
It might be hard to follow but if you look at the overall packet flow our traffic will come in from the physical in-interface on the left there and then go into the bridge from section A, in the bridging diagram it shows you how this traffic will flow, since the traffic is not coming from outside the router and is destined for an address not on this router this will not be an input chain instead since the traffic is originating from outside the router is is destined for an address not on this router it will be a forward chain packet.
so you can see in here that the traffic will not be processed by anything to do with the queues which is over in the routing diagram section, this will be regardless of hardware offloading as that is more about the process of if the traffic is handled by the CPU or by the switch chip.
Going back to the bridging diagram if you look there is a section saying use-ip-firewall? at these points if you enable this setting in the bridge the traffic is able to be processed by the prerouting, forward and postrouting chains of the IP firewall.
hopefully you can follow along with that explanation, I know im not the best at explaining so if you dont understand any part of this let me know and ill try and follow up with something better